Risk Prediction for Complications in Inflammatory Bowel Disease Surgery: External Validation of the American College of Surgeons’ National Surgical Quality Improvement Program Surgical Risk Calculator

Abstract

Abstract Background and Aims Many patients with inflammatory bowel disease [IBD] require surgery during their disease course. Having individual risk predictions available prior to surgery could aid in better informed decision making for personalised treatment trajectories in IBD surgery. The American College of Surgeons National Surgical Quality Improvement Program [ACS NSQIP] has developed a surgical risk calculator that calculates risks for postoperative outcomes using 20 patient and surgical predictors. We aimed to validate the calculator for IBD surgery to determine its accuracy in this patient cohort. Methods Predicted risks were calculated for patients operated for IBD between December 2017 and January 2022 at two tertiary centres and compared with actual outcomes within 30 postoperative days. Predictive performance was assessed for several postoperative complications, using metrics for discrimination and calibration. Results Risks were calculated for 508 patient trajectories undergoing surgery for IBD. Incidence of any complication, serious complications, reoperation, and readmission were 32.1%, 21.1%, 15.2%, and 18.3%, respectively. Of 212 patients with an anastomosis, 19 experienced leakage [9.0%]. Discriminative performance and calibration were modest. Risk prediction for any complication, serious complication, reoperation, readmission, and anastomotic leakage had a c statistic of 0.605 (95% confidence interval [CI] 0.534-0.640), 0.623 [95% CI 0.558-0.688], 0.590 [95% CI 0.513-0.668], 0.621 [95% CI 0.557-0.685], and 0.574 [95% CI 0.396-0.751], respectively, and a Brier score of 0.240, 0.166, 0.138, 0.152, and 0.113, respectively. Conclusions The accuracy of risks calculated by the ACS NSQIP Surgical Risk Calculator was deemed insufficient for patients undergoing surgery for IBD, generally underestimating postoperative risks. Recalibration or additional variables could be necessary to predict risks in this cohort.